The first principle: every number ties back to a GL line

The reporting most operators inherit is something between a marketing artifact and a forensic puzzle. The slide deck has numbers that don’t tie to the dashboard; the dashboard has numbers that don’t tie to QuickBooks; QuickBooks has numbers that don’t tie to the bank. By the time the board asks a follow-up question, half the meeting is spent reconciling which number is the right number.

Every report this engagement produces has the same property: any line on any view can be traced back to a specific account, class, and period in the GL, and the back-traced number matches. That’s the load-bearing rule. The dashboard isn’t a separate truth from the books; it’s a different presentation of the same truth. The board packet isn’t a hand-built deck; it’s an export.

The discipline this requires upstream is real and is the reason this line works best on top of §3 (Bookkeeping). Reporting on top of bookkeeping someone else owns produces an inheritance dispute every month — when the report and the books disagree, whose number is right? Same firm, same numbers, same close, and the question doesn’t come up.

The monthly board packet

Standard packet is P&L, balance sheet, cash flow, all on the accrual basis with comparatives — month-over-month, year-over-year, and against budget if there’s an active plan. Every line has a footnote when it moved materially, written by the CPA who closed the books, not generated. The packet ships as a PDF and is also available as a live view in whatever BI tool the engagement is on — Metabase by default, Looker Studio for clients who want Google-stack integration, Sigma for the ones that need spreadsheet-shaped exploration, custom for the rare case where the standard tools don’t fit.

The packet also includes an operator commentary page — one page, plain English, what happened in the month and what to watch in the next. We’ve found this to be the page most operators actually read first. The numbers tell you what happened; the commentary tells you what to do with it. Writing the commentary is part of the close.

What’s not in the standard packet: vanity metrics, gradient backgrounds, slideware-shaped narrative. Operators who’ve been around a few iterations recognize the difference between reports designed to look impressive and reports designed to be used; the packet is built for the second case.

The 13-week rolling cash forecast

Of every artifact this engagement produces, the 13-week cash forecast is the one operators reference most often. It’s a weekly refresh, by week, of expected receipts and disbursements — AR collections by aging cohort, AP payments by due date, payroll by cycle, recurring revenue, debt service, taxes, the seasonal patterns extracted from the trailing 24 months.

The forecast is built bottom-up from the GL, not top-down from a budget. That means it changes when the underlying drivers change — a slipped collection moves the line, a hire moves the line, a renewed contract moves the line. It’s not a strategic instrument; it’s an operating one. The point isn’t to predict the year. The point is to know on Monday morning whether you can fund payroll on Friday three weeks out.

Methodology, briefly: each line is its own model — collections lag the invoice by a distribution we fit per customer cohort, AP runs on terms net of the actual payment-behavior pattern, payroll runs on the calendar, recurring revenue runs on the contract base. The forecast is the sum, with a confidence interval on each line that widens as the horizon grows. By week 13 the interval is wide enough that the number is directional, not precise. By week 4 it’s tight enough to actually plan against.

The KPI catalogue

Every business has a different set of KPIs that matter, and there’s no formula for which ones do. The catalogue is the result of a conversation — usually a couple of hours during onboarding, sometimes more — to identify the half-dozen numbers the operator wants on a dashboard daily, the dozen-or-so they want monthly, and the rare aggregate metrics they want quarterly.

For each KPI we write a definition: source query, GL accounts and classes that contribute, exclusions, time window, refresh cadence, owner. The definition lives with the dashboard. When someone six months from now asks why the gross-margin number on the dashboard is different from the gross-margin number on the P&L, the answer is in the definition file, not in an inherited mystery.

A KPI without an owner is a KPI nobody trusts. We assign one for every metric on the catalogue. The owner is the operator who has the authority to act on the number, not the analyst who built the query.

On Tableau

The page lists Tableau as “we don’t take new Tableau seats — ask.” The reason is operational. Tableau is excellent software. It’s also expensive to operate at the scale most of the roster runs at, the licensing model penalizes the kind of read-only operator access that makes a dashboard useful, and the development cycle on a Tableau workbook is slower than the equivalent on Metabase or Sigma for the same chart. If an engagement already runs on Tableau and the existing investment is significant, we’ll happily continue on it. For new engagements, the recommendation is one of the alternatives in the tools list.

Multi-currency

Multi-currency consolidation is excluded from the standard engagement because the work changes shape — translation rules, intercompany hedging, FX gain/loss segregation, reporting-currency selection — and the right scope is a separate conversation, not a bullet in the retainer. We do this work, frequently, but as its own line. If a business is single-currency today and the operator is asking the multi-currency question because of a planned expansion, the right answer is usually “let’s scope this when the expansion is six months out, not now.”